Temporal and Spatial Patterns in Distemper Virus Cases Reported to SCWDS 1975-2013

Summary

Canine distemper is an important infectious disease that affects many mammal species. There is evidence of CDV infection in all terrestrial carnivores families and some marine carnivore families. CDV has been detected in wild animal species such as the African lion and Amur tigers and has been responsible for substantial population declines of the animals during outbreaks. Analysis was conducted on passive surveiilance necropsy diagnosed wild animal cases presneted to the South-Eastern Cooperative Wildlife Disease Study between 1975 and 2013. These analyses demonstrated the ability to use an ARIMA model to predict the numbers of Raccoon cases in a given month using data on the number of Gray Foxes cases from the previous three months. There also appears to be temporal trends in cases in both species, with cases more likely to occur during the breeding season. Spatially there tends to be clustering of cases of both species within the same areas, with cases appearing more likely to occur in area of medium human population density with fewer cases in very densely and sparsely populated areas, which may correspond to suburban areas where there is likely more contact with domestic dogs.

Introduction

Canine distemper is an important infectious disease that affects many mammal species. It is recognized as having the second highest case fatality rate of canine diseases , behind rabies (???). The causative agent, canine distemper virus (CDV) is an enveloped, single stranded, negative sense RNA virus in the Morbillivirus family. The major route of transmission is through aerosolization of virus containing respiratory exudate (Deem, Spelman, Yates, & Montali, 2000). CDV is highly infectious and virus may be shed for 60-90 days post infection (Greene & Appel, 1990). Distemper pathogenesis is well characterized in domestic dogs, with infections in wild species following a similar pattern. There is often a systemic infection with viremia and variable central nervous system involvement, dependent on the host’s immune response. Animals with a strong immune response will clear the virus and remain asymptomatic. A weak response will result in systemic disease and severe morbidity, often ultimately being fatal. An intermediate or delayed response results in delayed neurological signs and hyperkeratotic footpad lesions- the origin of the colloquial term “hard pad disease” (Greene & Appel, 1990). The clinical signs of CDV infection are influenced by multiple factors, including; strain virulence, environmental conditions, host age and immune status and the animal species infected (Deem et al., 2000). CDV results in a highly immunizing, acute infection. This type of pathogen normally requires high densities and large populations of hosts for long‐term persistence. CDV has been shown to be able to survive in wild carnivores due to multihost transmission, which overcomes the obstacle of population size and density within species (Almberg, Cross, & Smith, 2010). There is evidence of CDV infection in all terrestrial carnivores families and some marine carnivore families (Deem et al., 2000). Morbidity and mortality varies depending on the species but closely resembles rabies in wild carnivores (Hoff, Bigler, Proctor, & Stallings, 1974). The Mustelidae family is among the species with the highest fatality rate, while the domestic dog can be a asymptomatic carrier (Deem et al., 2000). CDV has been detected in wild animal species such as the African lion and Amur tigers and has been responsible for substantial population declines of the animals during outbreaks (Roelke-Parker et al., 1996; Seimon et al., 2013). CDV is likely to have broad consequences for the health of both free-living and domestic carnivore species (???). CDV’s similarity to the measles virus suggest a shared ancestor and a zoonotic potential for CDV.

Canine Distemper Virus in the U.S.

CDV is seen most commonly in domestic cats and dogs, but frequent cross species transmission does occur in non-domestic carnivores (Greene & Appel, 1990). Severity in domestic dogs depends on the animals’ immune status and age in addition to strain virulence (Beineke, Baumgärtner, & Wohlsein, 2015). In the US, raccoons (Procyon lotor), foxes (Vulpes vulpes and Urocyon cinereoargenteus), coyotes (Canis latrans), wolves (Canis lupus) , skunks (Mephitis mephitis), badgers (Taxidea taxus), mink (Mustela vison) and ferrets (Mustelidae spp.) are among the species susceptible to CDV infection(Kapil et al., 2008). In domestic ferrets mortality rates can reach 100% (Deem et al., 2000). CDV has been responsible for population declines of endangered mustelids like the black-footed ferret (Williams, Thorne, Appel, & Belitsky, 1988). CDV is also endemic in the eastern U.S. raccoon population. Raccoons are thought to be a reservoir for other wild animals and domestic dogs as well as other species of carnivores (Roscoe, 1993). CDV has been found to persist in areas like Yellow Stone national park, which has a diverse carnivore population. Multiple outbreaks have occurred in the wolf, coyote and cougar populations (Almberg et al., 2010; Almberg, Mech, Smith, Sheldon, & Crabtree, 2009). It has been shown that Raccoons are able to maintain higher density populations within urban and suburban environments (???). It was also demonstrated that these higher density populations are more likely to suffer mortality due to infectious disease. As distemper is a density dependent disease, it is likely that this disease incidence is influenced by urban vs rural environments. Although raccoons are thought to be a major reservoir for CDV, little research has been done to identify the disease dynamics within this population. Available data is sparse, dated and focuses on individual states and discrete sites. CDV has also been shown to be a mjor cause of disease in the south eastern USS, with 78% of necropsy cases being diagnosed with the disease in one study (???). The objective of this study is to identify spatial and temporal patterns in CDV cases, in a range of wild mammals, reported to the Southeastern Cooperative Wildlife Disease Study from 1975 to 2013.

Description of data and data source

Data was recorded from Canine distemper positive cases submitted to the Southeastern Cooperative Wildlife Disease Study (SCWDS) between 1975 and 2013. Cases were identified as CDV by fluorescent antibody testing and/or histologic diagnosis of characteristic lesions. Species, date of submission, county of origin, and sex were noted. A total number of 701 positive cases were submitted from 13 states over the 38-year period. Positive cases were comprised of raccoons (n=462), gray foxes (n=211), striped skunks (n=13), coyotes (n=7), red foxes (n=3), gray wolves (n=3), one mink and a black bear.

Census and county land area data for Georgia was accessed and downloaded from census.gov.

Questions to be addressed

1.Are there temporal trends in cases diagnosed? These may be longer term trends or seasonal trends related to breeding seasons. I hypothesize that there will be increased likelihood of cases occurring during the breeding season of wild mammals compared to the rest of the year.

2.Are there patterns in the timing of species being diagnosed suggesting cross species infection? Raccoons are considered primary reservoirs, are peaks in raccoon cases followed by cases in other species suggesting spillover? I hypothesize that there will be peaks in numbers of Raccoon CDV cases that are followed by cases in other species.

3.Are there spatial patterns of infection within the southeast? These patterns may be related to human activities and population density. Prevalence of CDV is reported to be higher in suburban foxes (???) I expect that there will be a greater likelihood of cases in areas of higher human population density compared to lower density.

Methods

Data aquisition

Data of animals brought to SCWDS between 1975 and 2013, which were diagnosed as having CDV at post mortem. Cases were identified as CDV by fluorescent antibody testing and/or histologic diagnosis of characteristic lesions. Species, date of submission, county of origin, and sex were noted.

Overview of Data

CDV case data contained the follwing variables; Case number, State, county, area, Sex, Species, Age and collection year. Additional data incluing specific collection dates was also used from a seperate spread sheet for the time series analysis.

Data import and cleaning

Data was imported into R Studio and cleaned to correct data entry errors and missing data. Detailed description of data analysis and cleaning is contained in the processing scripts within the project repository in the processing_code subfolder.

Time Series analsys

time series analysis and ARIMA model construction was conducted using the “fpp2” package from Forecasting: principles and practice, Hyndman & Athanasopoulos.(Hyndman & Athanasopoulos, 2018) For training and testing of the ARIMA models for predicting Raccoon and Gray Fox cases per month, the sequence of months from April 1975-September 2013 (n=462) was divided at into the training set, up to n=368 ie. November 2005, and the testing set from n=369 to n=462, ie. December 2005-September 2013.

Results

Data Exploration

Figure 1:Number of cases of CDV per state, submitted to SCWDS, 1075-2013

Figure 2:CDV cases per state submitted to SCWDS, 1975-2013 .

A total of 16 states clustered in the south east were represented in the data set with Georgia containing the majority of cases (n=422).

Figure 3:Number of CDV cases per species submitted to SCWDS, 1975-2013

There are eight species represented, with Gray Foxes and Raccoons making up the majority of cases

Figure 5:Total number of cases of CDV diagnosed at SCWDS per Year, 1975-2013.

Initial probing of the data set revealed the vast majority of cases to be submitted from the state of Georgia. This is understandable as SCWDS is located in Athens, GA. The other feature is that almost all of the submitted cases are Raccoons or Gray Foxes. From this point, data exploration and analysis will focus only on Gray foxes and Raccoons in the state of Georgia as this compromises the majority of cases.

Bivariate analysis of Species and Age of CDV cases submitted to SCWDS, 1975-2013

The age data is difficult to use as it is a particularly subjective measure in this case and there would need to be very marked changes for any relationships to be suggested.

Bivariate analysis of Species and Sex of CDV cases submitted to SCWDS, 1975-2013

Raccoon and Gray Fox cases in Georgia presented to SCWDS, 1975-2013

Raccoon and Gray Fox cases per breeeding season adjusted

In this case the different parts of the reproduction cycle were defined as the Breeding season being February to April, Lactation as May to July and the rest of the year non-Breeding. The number of cases recorded in each of these parts of the reprodiuctive cyce were adjusted by dividing by the length of the segment 9iin minths to account for any differences in the nuber of cases due to the lenght of the recording period.

Spatial Mapping of Georgia Data

Further qualitative analysis was conducted by mapping disease presence over time at county level.

Total Raccoon and Gray Fox cases in Georgia presented to SCWDS, 1975-2013, per county

County Presence over Time of CDV, from cases submitted to SCWDS, 1975-2013

The presence of CDV postive necropsy cases in Raccoons or Gray Foxes in each county is shown fro the years 1975-2013. Presence is used in this scenario rather than number of cases, as with the epidemiology of the disease, one can assume that a symptomatic case corresponds to an outbreak, especially in this data set as the surveillance is passive so cases are likely to be missed. The majority of counties with postive cases appear to occur in the northern part of the state with another small cluster in the south-east.

This was further analysed by comparing this distribution of cases over time by splitting cases by species.

Cases of CDV in Gray Foxes and Raccoons in Georgia counties, from cases submitted to SCWDS, 1975-2013

Individual cases of Raccoon and Fox CDV are plotted over time. As cases were only recorded to the county level, the points are plotted using county cenrtroid coordinates and a jitter to offset overlap of points.

Analysis of the level of clustering of cases was done using Ripley’s K analysis. Ripley’s K looks at the level of randomness of spatial points. It analyses whether a point is more or less likely to be near another point than to a randomly plotted point. The null hypothesis is that the cases are completely random spatially and that one case is not more or less likely to a closer to another case than a random point. In this analysis the cases of CDV were shown to be significantly more likely to be near another case than randomly spread, disproving the null hypothesis.

Ripley’s K analysis of CDV cases in Raccoons and Gray Foxes in Georgia, from cases submitted to SCWDs, 1975-2013

In the Ripley’s K plot, the red line is the Poisson distribution of randomly spread points, with error margin. The black line represents the CDV case locations. As this line is above the random distribution line and outside of the error margins, it can be said that these cases are significantly more clustered than a random distribution.

Raccoon and Gray Fox cases in Georgia presented to SCWDS, 1975-2013, and human population density

Raccoon and Gray Fox cases per spm against human population density

Cases of Raccoons and Gray Foxes were plotted both on a map showing population density per county and a regression of overall cases per area for each county against population density per sqm, which suggests a hump shaped relationship.

Time series analysis

Time series analysis was conducted on the number of cases in Gray foxes and Raccoons for the whole state of Georgia in the period 1975-2013 and ARIMA models using lagged predictors constructed using the “fpp2” package from Forecating: principles and practice, Hyndman & Athanasopoulos.(Hyndman & Athanasopoulos, 2018)

Raccoon and Gray Fox cases per month 1975-2013

Predictive Model

The general model for predicting number of Gray Fox cases using an auto-regression model found using the previous month’s predictive error along with this month’s Raccoon cases as the most accurate model for prediction. The models tested used an ARIMA base, which uses the numbers of cases from the preceding months to predict futre case numbers along with lagged regressors, in this case the Raccoon case numbers for previous months. Predictors were considered significant if the coefficeint was more than two standard errors away from zero.

Gray Fox Prediction Models

Model	Predictors	AIC
Fit_1	Gray_Fox_Cases_lag1 + Gray_Fox_Error_lag1	832.43
Fit_2	Gray_Fox_Error_lag1 + Raccoon_Cases_lag0	824.1
Fit_3	Gray_Fox_Error_lag1 + Raccoon_Cases_lag0 + Raccoon_Case_lag1	825.71
Fit_4	Gray_Fox_Error_lag1 + Raccoon_Cases_lag0 + Raccoon_Case_lag1 + Raccoon_Case_lag2	826.2
Fit_5	Gray_Fox_Error_lag1 + Raccoon_Cases_lag0 + Raccoon_Case_lag1 + Raccoon_Case_lag2 + Raccoon_Case_lag3	827.88
Null	Null Model	857.26

The autoARIMA part of the analysis tests for the most significant predictors in the autoregression part (Number of Gray Fox Cases from previous months) and the moving average part (the predicitve error in the model in predicting Gray Fox cases in previous months) along with the differencing and prints coefficients of those predictors which are significant. Akaike’s Information Criterion (AIC) is used as a measure of whether new predictors improve the model. This measure is penalised by adding 2 to the value for each new predictor added, so for an added predictor to be useful, it needs to decrease the AIC by more than 2, as a lower AIC represents a better model. As shown above, all of the models are an improvement on the null model.

In the best performing model, Fit_1, differencing and the autoregression were not useful predictors, but the error in the model from one month previously was a useful predictor, so is included. This means that if there is a large change in the number of cases in the previous month, ie. a large model error, then that is likely to be followed by another large change. On top of that base model, the number of cases of Raccoons in the current months and previous months were used as predictors, with increasing numbers of months into the past used in subsequent models.

Gray Fox Prediction General Model

Gray Fox Prediction Model

The process was repeated for predicting Raccoon cases with the strongest model using the previous months Raccoon cases and predictve error in additon to the number of Gray Fox cases this month and in the previous two months.

Raccoon Prediction Models

Model	Predictors	AIC
Fit_6	Raccoon_Cases_lag1 + Raccoon_Error_lag1	1190.63
Fit_7	Raccoon_Cases_lag1 + Raccoon_Error_lag1 + Gray_Fox_Cases_lag0	1184.48
Fit_8	Raccoon_Cases_lag1 + Raccoon_Error_lag1 + Gray_Fox_Cases_lag0 + Gray_Fox_Cases_lag1	1185.28
Fit_9	Raccoon_Cases_lag1 + Raccoon_Error_lag1 + Gray_Fox_Cases_lag0 + Gray_Fox_Cases_lag1 + Gray_Fox_Cases_lag2	1178.13
Fit_10	Raccoon_Cases_lag1 + Raccoon_Error_lag1 + Gray_Fox_Cases_lag0 + Gray_Fox_Cases_lag1 + Gray_Fox_Cases_lag2 + Gray_Fox_Cases_lag3	1180.03
Null	Null Model	1291.88

The inclusion of the number of Gray Fox cases in the current month and in the previous two months as predictors resulted in the best performing model in the set of Raccoon prediction models.

Raccoon Prediction General Model

Raccoon Prediction Model

The best model from each situation; predicting Raccoon cases and predicting Gray fox cases, were tested using the later part of the data set, with the Root Mean Squared Error (RMSE) being used to evaluate model performance.

Testing Prediction Models

	RMSE
Fit2_train	0.7399489
Fit2_test	0.2942214
Fit9_train	1.1911218
Fit9_test	0.5596792

The first model, which uses the number of Raccoon cases from the current month along with the predictive error for Gray Fox cases from the previous month to predict Gray Fox cases in the current month is the more accurate model with an RMSE of 0.2942214 for the test data, meaning it is easier to predict Gray Fox cases from this data. The second model predicts the Raccoon cases using the previous months Raccoon data and model error in addition to Gray Fox cases from the current month and the previous two months. Whilst this model was still quite accurate at predicting the number of cases it had a slighlty higher RMSE at 0.5596792.

Discussion

Initial introduction of CDV into wild carnivores in the U.S. in 1960’s was through grey foxes and subsequent spread to raccoons. (Hoff et al., 1974). Outbreak in Raccoons in Berlin, Germany appears to have originated in foxes with transmission seeming to readily occur between the species. (Rentería-Solís et al., 2014)

The analysis of the CDV case data in Raccoons and Gray Foxes suggests a relationship between cases in the two species, with the ARIMA model produced for Gray Fox cases suggesting that monthly case numbers in Gray Foxes can quite accurately be predicted using the previous months Gray Fox error data and the number of Raccoon cases from this month. A second model for the prediction of Raccoon cases using the previous months raccoon data in addition to Gray Fox cases from the previous three months was also quite accurate at predicting Raccoon cases, albeit it had a slightly higher error than the other model. In both predicitve models the test data had a lower RMSE than the training data. This can be explained by the variability of the dta over time. As shown in Figure 7 the year on year variation in number of cases is higher prior to 1996. As the training data ran from the period April 1975-November 2005, the model had to cope with much more variability in this set of data than with the test data from December 2005-September 2013, which will negatively impact the accuracy. 368 septmeber 2013 462

The spatial analysis of the data showed significant clustering of cases, which is in line with the ecology of the virus which tends to have epizootics where it spreads quickly amongst nearby susceptible. The primary mode of CDV infection is aerosol inhalation, suggesting habitat overlap and contact (Hoff et al., 1974). There appeared to be a much greater number of cases in the northern part of the state. Again the possible reasons for this are diverse; reporting bias due to a range of possible reasons; as the northern part of the state is more densely populated by humans, closer proximity to SCWDS, more suitable habitat and a consequently greater population of susceptible raccoons and gray foxes. Interestingly, there may be a parabolic relationship between population density and number of cases, with very high densities and very low densities having lower number of cases. This suggests that suburban areas may be a hotspot for disease circulation with these areas being attractive to Raccoons particlualry due to the balance of available habitat and easy scavenging opportunites. This is of particualr relevance as the disease is suggested to spillback from raccoons into dogs (Kapil & Yeary, 2011).

The data suggests a correlation between breeding season and the number of cases reported, with positives more likely to occur during the breeding season. This could be due to a number of reasons; there may be more contact between individual animals as they search for mates promoting aerosol spread of virus . Contact structure of a raccoon population can significantly impact disease transmission with Raccoon contact networks being shown to change depending on the season (breeding vs. non breeding); Rabies has been shown to spread quickly in raccoons when introduced during the breeding season.) (Reynolds, Hirsch, Gehrt, & Craft, 2015). Other potential influences of the breeding season and cases aside from contact may be the phycological strain of reproduction may leave animals more susceptible to the virus or as these cases are from animals found dead it could be that the increased movement from the breeding season leads to more being killed in other ways, such as on roads, and they happen to also be CDV positive at necropsy.

There as significant limitations to using this data set to draw conclusions about CDV cases in wild animals in Georgia. This data set comes from passive surveillance only, with it being reliant on dead animals being found and sent to SCWDS by DNR officers for necropsy. This leaves the data open to significant influence by reporting bias, as the distribution of DNR officers is not uniform across the state. Passive surveillance is also only showing part of the picture, with the results being heavily skewed towards symptomatic cases. Gray foxes and raccoons in Tennessee (Nov. 2013 – Aug. 2014) were infected frequently, but passive surveillance only captures animals showing clinical signs. Does not account for asymptomatic cases with 55% of asymptomatic animals tested being positive in this study (Pope, Miller, Riley, Anis, & Wilkes, 2016). There is also the difficulties associated with the size of the data set when it comes to trying to look at spatial and temporal trends together. Currently it is only possible to have meaningful numbers for analysis by looking at spatial trends using all the data, but obviously these cases are occuring over a very long period of time, the opposite is true for looking at temporal trends whereby only meanigful numbers can be gained by using cases over a huge area the size of Georgia.

Conclusions

The results from this analysis suggest that there are temporal and spatial patterns of CDV cases with the two most commonly infected wild carnivores in Georgia; Raccoons and Gray Foxes. Time series analysis and ARIMA model construction suggests that numbers of Raccoon cases can be predicted using data on the number of Gray Foxes cases from the previous three months. There appears to be temporal trends in cases in both species, with cases more likely to occur during the breeding season. Spatially there tends to be clustering of cases of both species within the same areas, with cases appearing more likely to occur in area of medium human population density with fewer cases in very densely and sparsely populated areas. Ultimately, there is enough trends suggested from this passively coolected data which would support a worthwhile study, with comprehsive sampling of wild carnivores over time and space to try to elucidate spillover trendsbetween species, paricularly in suburban areas.

References

Almberg, E. S., Cross, P. C., & Smith, D. W. (2010). Persistence of canine distemper virus in the Greater Yellowstone Ecosystem’s carnivore community. Ecological Applications. https://doi.org/10.1890/09-1225.1

Almberg, E. S., Mech, L. D., Smith, D. W., Sheldon, J. W., & Crabtree, R. L. (2009). A serological survey of infectious disease in yellowstone national park’s canid community. PLoS ONE. https://doi.org/10.1371/journal.pone.0007042

Beineke, A., Baumgärtner, W., & Wohlsein, P. (2015). Cross-species transmission of canine distemper virus-an update. https://doi.org/10.1016/j.onehlt.2015.09.002

Deem, S. L., Spelman, L. H., Yates, R. A., & Montali, R. J. (2000). CANINE distemper in terrestrial carnivores: A review. Journal of Zoo and Wildlife Medicine. https://doi.org/10.1638/1042-7260(2000)031[0441:cditca]2.0.co;2

Greene, C. E., & Appel, M. J. G. (1990). Canine distemper (pp. 226–241).

Hoff, G. L., Bigler, W. J., Proctor, S. J., & Stallings, L. P. (1974). Epizootic of canine distemper virus infection among urban raccoons and gray foxes. Journal of Wildlife Diseases. https://doi.org/10.7589/0090-3558-10.4.423

Hyndman, R. J., & Athanasopoulos, G. (2018). Forecasting: Principles and practice. OTexts: Melbourne, Australia.

Kapil, S., Allison, R. W., Johnston, L., Murray, B. L., Holland, S., Meinkoth, J., & Johnson, B. (2008). Canine distemper virus strains circulating among north American dogs. Clinical and Vaccine Immunology. https://doi.org/10.1128/CVI.00005-08

Kapil, S., & Yeary, T. J. (2011). Canine Distemper Spillover in Domestic Dogs from Urban Wildlife. https://doi.org/10.1016/j.cvsm.2011.08.005

Pope, J. P., Miller, D. L., Riley, M. C., Anis, E., & Wilkes, R. P. (2016). Characterization of a novel Canine distemper virus causing disease in wildlife. Journal of Veterinary Diagnostic Investigation. https://doi.org/10.1177/1040638716656025

Rentería-Solís, Z., Förster, C., Aue, A., Wittstatt, U., Wibbelt, G., & König, M. (2014). Canine distemper outbreak in raccoons suggests pathogen interspecies transmission amongst alien and native carnivores in urban areas from Germany. Veterinary Microbiology. https://doi.org/10.1016/j.vetmic.2014.08.034

Reynolds, J. J. H., Hirsch, B. T., Gehrt, S. D., & Craft, M. E. (2015). Raccoon contact networks predict seasonal susceptibility to rabies outbreaks and limitations of vaccination. Journal of Animal Ecology. https://doi.org/10.1111/1365-2656.12422

Roelke-Parker, M. E., Munson, L., Packer, C., Kock, R., Cleaveland, S., Carpenter, M., … Appel, M. J. G. (1996). A canine distemper virus epidemic in Serengeti lions (Panthera leo). Nature. https://doi.org/10.1038/379441a0

Roscoe, D. E. (1993). Epizootiology of canine distemper in New Jersey raccoons. Journal of Wildlife Diseases. https://doi.org/10.7589/0090-3558-29.3.390

Seimon, T. A., Miquelle, D. G., Chang, T. Y., Newton, A. L., Korotkova, I., Ivanchuk, G., … McAloose, D. (2013). Canine distemper virus: An emerging disease in wild endangered Amur tigers (Panthera tigris altaica). mBio. https://doi.org/10.1128/mBio.00410-13

Williams, E. S., Thorne, E. T., Appel, M. J., & Belitsky, D. W. (1988). Canine distemper in black-footed ferrets (Mustela nigripes) from Wyoming. Journal of Wildlife Diseases. https://doi.org/10.7589/0090-3558-24.3.385